This AISEM series provides a unique opportunity for both new and returning students to interact with groups of the UW Artificial Intelligence community (including machine learning, bioinformatics, computer vision etc.) and hear about their research. Each seminar will feature a unique group. The seminar includes several short talks given by students in that group and a social time which the PI will be present. **Refreshments** are provided.

In this seminar, we will introduce Prof. Jerry Zhu’s lab. Specific topics studied in this lab include semi-supervised learning, Bayesian nonparametrics, machine learning against bullying, and computational models for cognitive science. Two members of this lab, Kwang-Sung Jun and Bryan R. Gibson will present their work.

Title: Learning from Human List Production
Speaker: Kwang-Sung Jun

Human list production is the process of spontaneously generating an ordered list of items in response to some input. It generates a form of non-iid data with important applications in cognitive psychology and machine learning. We propose asampling-with-discounted-replacement (SWDR) model for human list production. We discuss its relation to other sampling models, and provide a parameter estimation procedure for learning. Two real-world applications demonstrate the value of our model: (i) in verbal fluency, our estimated parameters align well with psychological factors thought to influence behaviors in healthy and brain-damaged humans; (ii) in feature volunteering, our model improves the accuracy of text classifiers trained by Generalized Expectation criteria by learning from feature-label pairs spontaneously produced by human teachers.

Title: Using Machine Learning to Understand and Influence Human Categorization Behavior
Speaker: Bryan R. Gibson

In both machine learning (ML) and cognitive psychology, categorization is considered a basic task commonly encountered by learning agents as studied in both fields. While a great deal of work in cognitive psychology has been applied to understanding human learning in supervised categorization, almost no work has been done previously to investigate the effects of both labeled and unlabeled data as in the semi-supervised paradigm.
I will describe several projects which use ML to both (1) better understand how labeled and unlabeled data affect human learners in categorization tasks as well as (2) attempt to influence the resulting behavior using ideas and techniques derived from ML.